代谢综合征不同模型的验证性因子分析

Evaluation and comparison of models of metabolic syndrome using confirmatory factor analysis

目的 用验证性因子分析（confirmatory factor analysis，CFA）来评价和比较代谢综合征（MS）的6种假设模型。方法研究资料来自北京地区中老年人群糖尿病、高血压和肾脏病流行病学调查，共1411人（男性645人，女性766人），对其体质测量、血糖、胰岛素、血脂、血压等指标进行因子分析。根据以往的统计研究结果及临床实践假设了6个模型，其分别为：（1）模型1是由MS和胰岛素抵抗（IR）、肥胖、血脂紊乱、高血压构成的二阶模型[其中MS为二阶因子；IR、肥胖、血脂紊乱、高血压为一阶因子；空腹胰岛素（FINS）、餐后胰岛素（PINS）、空腹血糖（FBG）、餐后血糖（PBG）、体重指数（BMI）、腰围、甘油三酯（TG）、高密度脂蛋白胆固醇（HDL-C）、收缩压和舒张压为观察指标]；（2）模型2是由“IR”、“肥胖”、“血脂紊乱”、“高血压”构成的四因子模型；（3）模型3是由腰围、TG／HDL-C、稳态模型评估胰岛素抵抗指数（HOMA-IR）、平均动脉压（MAP）等构成的单因子模型；（4）模型4由FINS、腰围、TG以及收缩压作为次级因素的单因子模型；（5）模型5由FBG、腰围、TG以及收缩压作为次级因素的单因子模型；（6）模型6是由FINS、腰围、HDL—C和收缩压作为次级因素的单因子模型。使用SAS软件应用最大似然法对各个模型进行CFA评价，比较各假设模型的拟合度。结果单因子模型的各项拟合参数均优于四因子模型及二阶因子模型的各项拟合参数。以FINS、腰围、HDL-C和收缩压为观察指标，MS作为影响因子的单因子模型（模型6）拟合度最优[相对拟合指数（CFI）为0．995，近似均方根误差（SRMR）为0．013，X^2=4．233，P=0．434]。结论本研究结果支持目前普遍采用的MS定义，同时提示可能存在一个受MS各组分影响的共同因子，为说明MS是一个独立的疾病提供了有力证据。

Objective To evaluate and compare six competing models of metabolic syndrome（MS） using confirmatory factor analysis （CFA）. Methods The study included 1 411 subjects （645 men and 766 women） during a survey of diabetes mellitus, hypertention and renal diseases in Beijing area. Factor analysis was performed using the anthropometric measures and variables such as plasma insulin, blood glucose, blood lipids and blood pressure. The six models were hypothesized according to the previous studies about the factor analysis on the MS and the evaluated models were as follows： （ 1 ） a two-order model with hypothesized MS and the secondary factors which included insulin resistance （IR）, obesity, dyslipidemia and hypertension, and the fasting insulin （FINS）, postprandial insulin （PINS）, fasting blood glucose （FBG）, postprandial blood glucose （PBG）, body mass index （BMI）, waist circumference （WC）, triglyceride （TG）, high density lipoprotein-cholesterol （HDL-C）, systolic blood pressure （SBP） and diastolic blood pressure considered as the observational index; （2） a four-factor model with IR, obesity, dyslipidemia and hypertension; （3） a one-factor model with four secondary modules such as WC,TG/HDL-C, homeostasis model assessment of insulin resistance index（HOMA-IR） and mean arterial pressure （MAP） ; （4） a one-factor model with four secondary modules which were FINS, WC, TG and SBP; （5） a one- factor model with four secondary modules with were FBG, WC, TG and SBP; （6） a one-factor model with four secondary modules which were FINS, WC, HDL-C and SBP. Several CFAs were performed using SAS software with maximum likelihood estimation, and the fit indices of the different models were compared. Results These onefactor models yielded significantly better data-model fit than the four-factor or the two-order mode type. And the model 6 with its underlying factors including FINS, WC, HDL and SBP was supposed to be the most plausible one among the competing models （comparative fit index was 0. 995, root mean square residual was 0. 013 ）.Conclusion The results supported the generally accepted definition of MS. There might be a common factor that could influence the different aspects of MS, thus providing an evidence that the MS is an independent desease.